The zero-state response accounts for system behavior due only to an external input. For a linear system described by ˙x(t) = Ax(t) + Bu(t), set initial conditions x(0) = 0, then solve for x(zero-state) by integrating state equations with respect to u(t). Separately, the zero-input response considers A = matrix dynamics with x(0) as the non-zero initial condition and u(t) = 0 leading to evolution based solely on initial states. These dual analysis responses are superimposed for complete system behavior, often during responses to a sudden change like a step input .

A matrix is nilpotent if a positive integer k exists such that A^k = 0. For A to be nilpotent of order 2, A^2 must result in the zero matrix. Implications include that the system has no structural modes that persist indefinitely; dynamics linked to each mode decay to zero, reflecting in unobservable or unreachable states. In practical terms, this reveals transient-only dynamics, crucial for control schemes where lingering state presence skews desired system response, impacting controller/observer design .

Poles of the transfer function G(s) reflect the system's natural response modes and stability, directly connected to eigenvalues of the system matrix A. Poles, obtained from denominator roots of G(s), equate to eigenvalues for SISO systems, analyzing their real/imaginary components reveals stability (negative real poles) and oscillatory behavior (complex poles). A significant deviation from expected relation suggests latent states or modes affecting dynamics beyond observable outputs, pivotal in designing stable and responsive control systems. A mismatch often points to model reduction or estimation errors .

The transfer function G(s) is derived by taking the Laplace transform of the state-space representation to find Y(s)/U(s), which involves the matrices A, B, and C in G(s) = C(sI - A)^-1B + D. The degree of the transfer function corresponds to the highest power of the denominator when written in polynomial form, which is the number of poles of the system. However, this degree may differ from the number of states if the system has unattainable or unobservable states that do not influence output dynamics. This discrepancy involves controllability and observability theory and structural characteristics of matrix connections .

When u(t) = 1 + 2e^-2t, the state equation ˙x(t) = Ax(t) + Bu(t) results in a solution where the transient component disappears as t approaches infinity, leaving a steady state or causing parts of the state to grow unbounded depending on the eigenvalues' real parts. For x(t) = [x1(t) x2(t)] as t → ∞, x1 approaches a constant, while x2 diverges if the system is unstable. The output y(t) also reflects this behavior, becoming predominantly influenced by the steady-state values of the states, computed as y(∞). The specific dynamic of each variable is tied to how A's eigenvalues dictate stability .

The state transition matrix Φ(t), significant in determining system state x(t) over time, is defined as e^At where A is the system matrix. It transforms initial conditions into state behavior x(t) = Φ(t)x(0). To compute e^At, methods like matrix diagonalization (eigen-decomposition), Laplace transform for Φ(s) = (sI-A)^-1, or series expansion, are applied, depending on A's properties. This matrix's accuracy is crucial for forecasting x(t) development and resolving control inputs' effects, impacting observability and stability assessments .

Stability of the state-space model can be determined by analyzing the eigenvalues of the system matrix A. For A = [4 1; 0 10], solve λI - A = 0 to find eigenvalues λ = 4 and λ = 10. Since stable systems require all eigenvalues to have negative real parts (or zero real parts for marginal stability), this system with positive eigenvalues indicates instability. Thus, no eigenvalue has a negative real part, confirming that even a small disturbance can cause the state variables to grow indefinitely, showcasing instability .

Using different basis sets like {B, AB, A^2B} for system representation facilitates simplifying control and observability analyses by transforming the state space. This form highlights system dynamic properties: controllable subspace identification aids design by focusing on directly influenceable states, and in observable forms, states distinctly affect output derivation, streamlining observer design. This structured representation reveals clearer geometric insight into state transformations relative to input-output relation, essential in advanced control strategies for efficacy in complex dynamical systems modeling .

To compute e^At using the Laplace transform, observe that L{e^At} = (sI - A)^(-1). For matrix A = [0 1; 0 -2], (sI - A) = [s 1; 0 s+2]. The inverse (sI - A)^(-1) leads to the expression involving a partial fraction decomposition. Alternatively, the Cayley-Hamilton (C-H) Theorem states that matrix A satisfies its own characteristic polynomial, which here is λ^2 + 2λ = 0. Using this, express e^At as a polynomial in A based on the characteristic polynomial, giving a solution derived from power series expansion terms based on e^0 and e^-2t factors .

To compute the eigenvalues of matrix A = [0 1; 0 -2], we solve the characteristic equation det(A - λI) = 0, where I is the identity matrix. This results in the equation (0-λ)(-2-λ) - (1)(0) = λ^2 + 2λ. Solving λ^2+2λ = 0 gives λ = 0, -2. To find the eigenvectors, for λ = 0, solve (A - 0I)x = 0, leading to the eigenvector associated with λ = 0 as any scalar multiple of vector [1; 0]. For λ = -2, solve (A + 2I)x = 0 leading to the eigenvector as any scalar multiple of vector [1; -2].